A multi-lingual intelligent guide system and method based on a WeChat applet

Abstract

The application relates to the technical field of intelligent travel and tourism guide, and discloses a multilingual intelligent guide system and method based on a WeChat applet, which comprises a content update management module, a content consistency checking module, an AI question and answer index synchronization module, a content version package generation module, a code scanning access control module, a cache synchronization control module and an AI question and answer output module. After the system receives a content update request of a target exhibit, a candidate version of the exhibit content is generated, semantic unit matching and key entity matching are performed on a source language explanation text and a target language explanation text, and a content consistency evaluation result is obtained; when the release condition is met, an AI question and answer knowledge segment and an exhibit content version package are generated, and are set as the current effective version. By making the WeChat applet page display, voice playing and AI question and answer retrieval all based on the same version resource, the consistency of the guide content version is improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent cultural tourism guide technology, specifically a multilingual intelligent guide system and method based on WeChat mini-programs. Background Technology

[0002] With the development of digitalization in cultural and tourism venues such as museums, exhibition halls, memorial halls, and scenic spots, mobile-based intelligent tour guide systems are gradually replacing traditional audio guide machines. Existing tour guide systems typically access cloud-based explanations via QR codes, location recognition, or exhibit numbers, providing users with services such as text explanations, audio playback, multilingual switching, backend content management, and intelligent Q&A. Due to the advantages of WeChat mini-programs—no installation required, ease of use, and easy promotion—they are also increasingly being applied to venue tours, exhibit introductions, and visitor interactive Q&A scenarios.

[0003] However, existing mobile or WeChat-based tour guide systems typically only replace text, audio, images, or knowledge base content when updating exhibit content. They lack a unified version constraint mechanism for multilingual text, audio resources, AI Q&A knowledge snippets, and local cached resources within the mini-program for the same exhibit. When the exhibition hall modifies the exhibit's era, relationships between figures, its location, or the explanation content, issues easily arise such as updated Chinese text but outdated foreign language text, new content displayed on the page but old audio, AI Q&A calling historical knowledge snippets, and the mini-program continuing to load old cached content. This leads to inconsistencies in the content of the same exhibit across different languages, resource types, and interaction scenarios, affecting the accuracy, authority, and reliability of the tour guide content. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a multilingual intelligent tour guide system and method based on WeChat mini-programs, which solves the problem that existing tour guide systems are prone to version inconsistencies in multilingual text, voice resources, AI Q&A content, and mini-program cached content after exhibit content is updated.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a multilingual intelligent navigation system based on WeChat mini-programs, comprising: The content update management module is used to receive the exhibit content update request of the target exhibit and generate exhibit content candidate versions. The exhibit content candidate versions include exhibit identifier, source language explanation text, target language explanation text, audio resources, image resources and extended knowledge text. The content consistency verification module is used to perform semantic unit matching and key entity matching on the source language explanation text and the target language explanation text to generate a content consistency evaluation result. The AI ​​Q&A index synchronization module is used to generate AI Q&A knowledge fragments based on the candidate versions of the exhibit content when the content consistency evaluation results meet the release conditions, and to bind exhibit identifiers, language identifiers and candidate version identifiers to the AI ​​Q&A knowledge fragments. The content version package generation module is used to generate an exhibit content version package based on the candidate versions of the exhibit content, and set the exhibit content version package as the currently effective version. The exhibit content version package includes exhibit identifier, language identifier, audio resources, image resources, AI question and answer knowledge fragments, resource hash values, and scope of applicable permissions. The QR code access control module is used to respond to the WeChat mini program's request to scan the exhibit's QR code, obtain the exhibit identifier corresponding to the exhibit's QR code, and receive the language identifier, local cache version number, and permission token uploaded by the WeChat mini program. The cache synchronization control module is used to compare the local cache version number with the content main version number of the currently effective version, and when the local cache version number is inconsistent with the content main version number, it sends a cache invalidation instruction and navigation resources to the WeChat Mini Program. The AI ​​Q&A output module is used to respond to AI Q&A requests sent by WeChat Mini Programs. It limits the scope of AI Q&A knowledge fragment retrieval based on the exhibit identifier, language identifier, current effective version, and permission token, and generates guide answer content based on the retrieved AI Q&A knowledge fragments.

[0006] Preferably, the QR code access control module is further used to parse the QR code encoding information in the exhibit's QR code, the QR code encoding information including the exhibition hall identifier, exhibit identifier, and QR code batch identifier; Determine whether the exhibit's QR code belongs to a valid QR code batch based on the batch identifier of the QR code; If it belongs to a valid QR code batch, then query the currently effective version based on the exhibit identifier; If it does not belong to a valid QR code batch, a QR code expiration message or exhibit redirection information will be generated.

[0007] Preferably, the content consistency verification module includes: The semantic unit matching unit is used to split the source language explanation text into source language semantic units, split the target language explanation text into target language semantic units, and establish a semantic matching relationship between the source language semantic units and the target language semantic units. The key entity matching unit is used to extract key entities from the source language explanation text and the target language explanation text, and to establish entity matching relationships between key entities in the source language and key entities in the target language. The resource reference detection unit is used to detect the integrity of resource references among the target language explanatory text, audio resources, image resources, and extended knowledge text; The consistency evaluation unit is used to generate the content consistency evaluation result based on the semantic matching relationship, entity matching relationship, and resource reference integrity.

[0008] Preferably, the key entities include exhibit name, person name, historical period, place name, dynasty, artifact type and quantity information; The key entity matching unit is also used to standardize the key entities, generate key entity fingerprints, and determine whether the source language explanation text and the target language explanation text correspond to the same exhibit semantic version based on the key entity fingerprints.

[0009] Preferably, the content version package generation module is also used to generate a main content version number, a text resource version number, a voice resource version number, an image resource version number, and an AI question-and-answer knowledge fragment version number; Write the main version number of the content, the version number of the text resource, the version number of the voice resource, the version number of the image resource, the version number of the AI ​​question-and-answer knowledge fragment, the resource hash value, the key entity fingerprint, and the scope of permission application into the same exhibit content version package.

[0010] Preferably, the AI ​​question-answering index synchronization module is further used to segment the source language explanation text, the target language explanation text, and the extended knowledge text into candidate knowledge fragments; Semantic deduplication and entity completion are performed on the candidate knowledge fragments to obtain AI question-answering knowledge fragments; An AI question-and-answer retrieval index is generated based on the AI ​​question-and-answer knowledge fragments, and the AI ​​question-and-answer knowledge fragments corresponding to historical versions are marked as not currently in effect.

[0011] Preferably, the cache synchronization control module is also used to obtain the hash value of local resources uploaded by the WeChat mini program; The local resource hash value is compared with the resource hash value in the currently effective version to determine the differing resources; If the differential resources meet the differential update conditions, then a differential resource package is sent to the WeChat mini program; If the differential resources do not meet the differential update conditions, a complete resource package will be sent to the WeChat mini program.

[0012] Preferably, the content update management module is further configured to generate corresponding language audio resources based on the target language explanation text when the target language explanation text meets the publishing conditions, or to receive manually recorded audio resources corresponding to the target language explanation text; Calculate the resource hash value of the audio resource and bind the resource hash value of the audio resource to the text resource version number of the target language explanation text; When the version number of the text resource for the target language explanation text changes, the audio resource bound to the previous text resource version number is marked as pending update.

[0013] A multilingual intelligent navigation method based on WeChat Mini Programs, used in the aforementioned multilingual intelligent navigation system based on WeChat Mini Programs, the method comprising: S1. Receive the exhibit content update request for the target exhibit and generate candidate versions of the exhibit content; S2. Perform semantic unit matching and key entity matching on the source language explanation text and the target language explanation text to generate content consistency evaluation results; S3. When the content consistency evaluation result meets the release conditions, generate an AI question-and-answer knowledge fragment and bind the exhibit identifier, language identifier and candidate version identifier to the AI ​​question-and-answer knowledge fragment; S4. Generate an exhibit content version package based on the candidate versions of the exhibit content, and set the exhibit content version package as the currently effective version; S5. Respond to the WeChat mini program's request to scan the exhibit's QR code, and obtain the exhibit identifier, language identifier, local cache version number, and permission token; S6. Compare the local cache version number with the content main version number of the currently effective version, and issue a cache invalidation command and navigation resources when the local cache version number is inconsistent with the content main version number; S7. Responding to the AI ​​Q&A request sent by the WeChat mini program, the scope of AI Q&A knowledge fragment retrieval is limited based on the exhibit identifier, language identifier, current effective version and permission token, and guide answer content is generated.

[0014] The present invention also provides an electronic device, including a processor, a memory, and a communication interface. The processor, the memory, and the communication interface are connected through a communication bus. The memory stores a computer program, and when the computer program is executed by the processor, it implements the above-mentioned multilingual intelligent navigation method based on WeChat mini-program.

[0015] This invention provides a multilingual intelligent navigation system and method based on WeChat mini-programs. It has the following beneficial effects: 1. This invention binds the source language explanation text, target language explanation text, audio resources, image resources, AI Q&A knowledge fragments, resource hash values, and permission application scope to the same currently effective version through content consistency verification, AI Q&A index synchronization, and exhibit content version package generation. This ensures that the WeChat mini-program page display, audio playback, and AI Q&A retrieval are all based on the same version of resources. This reduces issues such as inconsistencies in multilingual text, audio-text asynchrony, AI Q&A referencing historical knowledge fragments, and the mini-program continuing to load old caches after exhibit content updates, thereby improving the version consistency and release reliability of the guide content.

[0016] 2. Before the release of candidate versions of exhibit content, this invention performs semantic unit matching and key entity matching on the source language explanation text and the target language explanation text, and generates content consistency evaluation results in combination with resource reference integrity. It can detect inconsistencies in key content such as historical dates, names of people, place names, types of artifacts, and quantity information in different languages ​​before release, reducing the workload of manually checking multilingual explanation content line by line and improving the efficiency of multilingual guide content review.

[0017] 3. This invention binds exhibit identifiers, language identifiers, candidate version identifiers, content main version numbers, and scope of permissions to AI Q&A knowledge fragments through an AI Q&A index synchronization module, and marks AI Q&A knowledge fragments corresponding to historical versions as not currently effective. This ensures that the AI ​​Q&A output module can only retrieve knowledge fragments within the scope limited by exhibit identifiers, language identifiers, currently effective versions, and permission tokens. This reduces the occurrence of AI Q&A calling old version knowledge, unapproved knowledge, or knowledge without permission, and improves the controllability of AI guide answers.

[0018] 4. This invention compares the local cache version number with the main version number of the currently effective version through the cache synchronization control module, and further determines the difference resources through the local resource hash value and the resource hash value in the currently effective version. This can ensure that the navigation content on the mini program is updated in a timely manner, and select to send differential resource packages or complete resource packages according to the difference resources, thereby reducing duplicate resource transmission and improving the resource loading efficiency after scanning the code. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the system framework of a multilingual intelligent navigation system based on WeChat Mini Programs according to the present invention. Figure 2 This is a flowchart illustrating a multilingual intelligent navigation method based on WeChat Mini Programs according to the present invention. Detailed Implementation

[0020] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1 This embodiment provides a multilingual intelligent navigation system based on WeChat mini-programs, such as... Figure 1 As shown, it includes a content update management module, a content consistency verification module, an AI question and answer index synchronization module, a content version package generation module, a QR code access control module, a cache synchronization control module, and an AI question and answer output module.

[0022] In this system, WeChat Mini Programs serve as the user access terminal, while cloud servers act as the content processing terminal. The cloud servers can connect to a content database, a resource repository, an AI question-and-answer retrieval library, and a permission verification service. The content database stores exhibit identifiers, source language explanation text, target language explanation text, extended knowledge text, language identifiers, candidate version identifiers, and the currently active version. The resource repository stores audio resources, image resources, and the corresponding resource hash values. The AI ​​question-and-answer retrieval library stores AI question-and-answer knowledge fragments and their indexes. The permission verification service verifies permission tokens.

[0023] The content update management module is used to receive content update requests for target exhibits and generate candidate versions of exhibit content. The candidate versions of exhibit content include exhibit identifiers, source language explanatory text, target language explanatory text, audio resources, image resources, and extended knowledge text.

[0024] Specifically, the target exhibits are physical exhibits, digital exhibits, temporary exhibits, or exhibition units within the exhibition hall that require guided tour services. Backend administrators can input or modify the source language explanation text, target language explanation text, audio resources, image resources, and extended knowledge text for the target exhibits through a cloud-based management page. The source language explanation text can be in Chinese, and the target language explanation text can be in one or more foreign languages ​​such as English, Japanese, Korean, French, or Spanish.

[0025] Upon receiving a request to update exhibit content, the content update management module does not directly publish the updated content to the WeChat mini-program. Instead, it first generates candidate versions of the exhibit content. Each candidate version has a candidate version identifier and is marked as pending verification. The candidate version identifier can be generated by combining the exhibit identifier, language identifier, update time, and an incrementing number, or it can be automatically generated by the cloud server according to version management rules.

[0026] Furthermore, the content update management module is also used to generate corresponding language audio resources based on the target language explanatory text when the target language explanatory text meets the publication conditions, or to receive manually recorded audio resources corresponding to the target language explanatory text. The system calculates the resource hash value of the audio resource and binds the resource hash value of the audio resource to the text resource version number of the target language explanatory text. When the text resource version number of the target language explanatory text changes, the audio resource bound to the previous text resource version number is marked as pending update.

[0027] Specifically, audio resources can be generated through speech synthesis services or recorded and uploaded by the narrator. The resource hash value of the audio resource can be calculated using MD5, SHA-1, SHA-256, or other hash algorithms. By binding the resource hash value of the audio resource to the text resource version number of the target language narration text, the problem of the audio resource still using the old version when the target language narration text has been updated can be avoided.

[0028] The content consistency verification module is used to perform semantic unit matching and key entity matching between the source language explanation text and the target language explanation text, and generate content consistency evaluation results.

[0029] Furthermore, the content consistency verification module includes a semantic unit matching unit, a key entity matching unit, a resource reference detection unit, and a consistency evaluation unit.

[0030] The semantic unit matching unit is used to split the source language explanation text into source language semantic units, split the target language explanation text into target language semantic units, and establish semantic matching relationships between source language semantic units and target language semantic units.

[0031] Specifically, semantic units can be sentences, phrases, paragraphs, or text fragments segmented according to punctuation and semantic boundaries. The system segments the source language and target language explanatory texts into sentences, removes redundant spaces, standardizes punctuation, standardizes terminology, and filters stop words. Then, it inputs each semantic unit into a cross-language semantic vector model to obtain source language semantic vectors and target language semantic vectors. The cross-language semantic vector model can employ a multilingual BERT model, a cross-language sentence vector model, or a semantic matching model fine-tuned from the exhibition corpus.

[0032] The semantic unit matching unit calculates the similarity between the source language semantic vector and the target language semantic vector, and establishes semantic matching relationships according to the similarity. If a source language semantic unit cannot be matched with a target language semantic unit whose similarity reaches a preset threshold, the source language semantic unit is marked as an uncovered semantic unit; if a target language semantic unit has high similarity with multiple source language semantic units, it is marked as a semantic unit awaiting manual review.

[0033] The key entity matching unit is used to extract key entities from the source language explanatory text and the target language explanatory text, and to establish entity matching relationships between key entities in the source language and key entities in the target language.

[0034] Specifically, key entities include exhibit names, personal names, historical dates, place names, dynasties, artifact types, and quantity information. The key entity matching unit can extract key entities using a named entity recognition model, an exhibit terminology dictionary, a historical date rule base, and a manually maintained thesaurus. For example, the system can standardize "late Qingdynasty" (Late Qingperiod) as the same historical date entity; and standardize "blue and white porcelain vase" and "blue-and-white porcelain vase" as the same artifact type entity.

[0035] The key entity matching unit is also used to standardize key entities, generate key entity fingerprints, and determine whether the source language explanatory text and the target language explanatory text correspond to the same semantic version of the exhibit based on the key entity fingerprints. Key entity fingerprints can be obtained by hashing the standardized exhibit name, person name, historical date, place name, dynasty, artifact type, and quantity information in a preset order, or they can be directly composed of a set of standardized entities.

[0036] The resource reference detection unit is used to detect the integrity of resource references among the target language explanatory text, audio resources, image resources, and extended knowledge text.

[0037] Specifically, the resource reference detection unit checks whether the image ID, audio ID, exhibit ID, and extended knowledge ID referenced in the target language explanatory text exist. It also checks whether the audio resource is bound to the text resource version number of the target language explanatory text, whether the image resource belongs to the same exhibit identifier, and whether the extended knowledge text is associated with the current candidate version identifier. If the target language explanatory text references a non-existent image resource, or if the audio resource is bound to the previous text resource version number, the resource reference detection unit generates a resource reference anomaly flag.

[0038] The consistency evaluation unit is used to generate content consistency evaluation results based on semantic matching relationships, entity matching relationships, and resource reference integrity.

[0039] Specifically, the consistency evaluation unit determines the semantic unit matching score, key entity matching score, and resource reference integrity score. The semantic unit matching score is determined based on the degree of matching coverage between source language semantic units and target language semantic units, the number of uncovered semantic units, and the number of semantic units requiring manual review. The key entity matching score is determined based on whether the key entity fingerprints are consistent, the number of missing key entities, and the number of conflicting key entities. The resource reference integrity score is determined based on the reference integrity among the target language explanatory text, audio resources, image resources, and extended knowledge text.

[0041] When the content consistency evaluation result reaches the release threshold and there are no key entity conflicts, the content consistency verification module determines that the candidate version of the exhibit content meets the release conditions. When the content consistency evaluation result does not reach the release threshold, or there are key entity conflicts, the content consistency verification module marks the candidate version of the exhibit content as pending correction and generates a correction prompt. The correction prompt may include uncovered semantic units, conflicting key entities, missing audio resources, invalid image resources, and abnormal extended knowledge text.

[0042] The AI ​​Q&A index synchronization module is used to generate AI Q&A knowledge fragments based on candidate versions of exhibit content when the content consistency evaluation results meet the release conditions, and to bind exhibit identifiers, language identifiers and candidate version identifiers to the AI ​​Q&A knowledge fragments.

[0043] Specifically, the AI ​​question-answering index synchronization module segments the source language explanation text, target language explanation text, and extended knowledge text into candidate knowledge fragments. Candidate knowledge fragments can be segmented according to paragraphs, semantic topics, entity categories, or preset lengths. Each candidate knowledge fragment includes at least the fragment text, exhibit identifier, language identifier, and candidate version identifier.

[0044] Furthermore, the AI ​​question-answering index synchronization module is also used to perform semantic deduplication and entity completion on candidate knowledge fragments to obtain AI question-answering knowledge fragments. Semantic deduplication is used to delete candidate knowledge fragments with duplicate meanings in candidate versions of the same exhibit content, while entity completion is used to fill in the omitted exhibit names, historical periods, figures, or artifact types in the candidate knowledge fragments, so as to avoid incomplete answers due to missing context during AI question-answering retrieval.

[0045] For example, if a candidate knowledge fragment only records that its patterns reflect the aesthetic trends of the time, the AI ​​question-answering index synchronization module can complete it by using the exhibit identifier and key entity fingerprint to which the candidate knowledge fragment belongs, to show that the patterns of the target exhibit reflect the aesthetic trends of the late Qing Dynasty. The completed knowledge fragment does not change the original factual content, but adds the contextual entities needed for retrieval and answering.

[0046] The AI ​​Q&A index synchronization module generates an AI Q&A retrieval index based on AI Q&A knowledge fragments. The AI ​​Q&A retrieval index can include a vector index, a keyword inverted index, or a combination of both. Each AI Q&A knowledge fragment is written with an exhibition hall identifier, exhibit identifier, language identifier, candidate version identifier, main content version number, and applicable scope of permissions. When a new exhibit content version package is set as the currently effective version, the AI ​​Q&A knowledge fragments corresponding to historical versions are marked as not currently effective.

[0047] The content version package generation module is used to generate an exhibit content version package based on the candidate versions of exhibit content, and set the exhibit content version package as the currently effective version. The exhibit content version package includes exhibit identifier, language identifier, audio resources, image resources, AI question and answer knowledge fragments, resource hash values, and scope of applicable permissions.

[0048] Furthermore, the content version package generation module is also used to generate the main content version number, text resource version number, voice resource version number, image resource version number, and AI question-and-answer knowledge fragment version number; and to write the main content version number, text resource version number, voice resource version number, image resource version number, AI question-and-answer knowledge fragment version number, resource hash value, key entity fingerprint, and scope of permission application into the same exhibit content version package.

[0049] Specifically, the main content version number indicates the overall content version of the target exhibit at a specific release point. The text resource version number indicates the text version of the source language explanation text or the target language explanation text. The audio resource version number indicates the audio resource version associated with the text resource version number. The image resource version number indicates the version of the image resource collection. The AI ​​question-answering knowledge fragment version number indicates the version of the AI ​​question-answering knowledge fragment collection under the current main content version number.

[0050] Resource hash values ​​can include text resource hash values, voice resource hash values, image resource hash values, and AI question-and-answer knowledge fragment hash values. The scope of access permissions indicates the types of users, exhibition halls, exhibits, languages, and AI question-and-answer functionality that can access this exhibit content package.

[0051] During deployment, the content version package generation module employs an atomic deployment method, setting all validated exhibit content version packages as the currently effective version in a single deployment transaction. Atomic deployment means that text resources, audio resources, image resources, and AI question-and-answer knowledge fragments take effect within the same deployment transaction; if any resource fails to be written or validated, the exhibit content version package will not be set as the currently effective version. Older versions of content can be retained in the historical version repository and marked as historical versions for later tracing or rollback.

[0052] The QR code access control module is used to respond to the WeChat mini program's request to scan the exhibit's QR code, obtain the exhibit identifier corresponding to the exhibit's QR code, and receive the language identifier, local cache version number, and permission token uploaded by the WeChat mini program.

[0053] Specifically, exhibit QR codes can be placed next to the target exhibit, on the display case label, or on the exhibition hall guide sign. After scanning the exhibit QR code with a WeChat mini-program, the QR code access control module parses the QR code encoding information. The QR code encoding information includes the exhibition hall identifier, the exhibit identifier, and the QR code batch identifier. The exhibition hall identifier is used to identify the exhibition hall to which the target exhibit belongs, the exhibit identifier is used to identify the target exhibit, and the QR code batch identifier is used to determine whether the QR code belongs to the current valid QR code batch.

[0054] The QR code access control module determines whether an exhibit's QR code belongs to a valid batch based on the batch identifier. If it does, it queries the currently active version based on the exhibit identifier; otherwise, it generates a QR code expiration message or exhibit redirection information. The exhibit redirection information can be used to guide the WeChat mini-program to access the new exhibit content version package after the exhibit is moved to a mobile exhibition hall or the QR code is changed.

[0055] When sending a QR code scanning request, the WeChat Mini Program also uploads a language identifier, a local cache version number, and a permission token. The language identifier indicates the language currently selected by the user for the tour; the local cache version number indicates the version of the exhibit content package that has been cached locally by the WeChat Mini Program; and the permission token indicates the user's access permissions to the exhibition hall, the target exhibit, the language, and the AI ​​Q&A function.

[0056] In another implementation, the WeChat mini program can also generate a guided tour trigger request based on GPS location information and near-field beacon trigger information. Specifically, the WeChat mini program obtains the user's terminal's GPS location information and receives near-field beacon trigger information sent by near-field beacons within the target exhibition hall. The cloud server determines whether the user has entered the target exhibition hall or area based on the GPS location information and identifies the target exhibit the user is approaching based on the near-field beacon trigger information.

[0057] When GPS location information indicates that the user is within the target exhibition area, and the near-field beacon trigger information is bound to the exhibit's identifier, the QR code access control module identifies the target exhibit as the exhibit to be explained and obtains the corresponding exhibit identifier. The WeChat mini-program simultaneously uploads the language identifier, local cache version number, and permission token. Based on the exhibit identifier, language identifier, local cache version number, and permission token, the system continues to perform current version query, cache synchronization, and guide resource distribution.

[0058] Therefore, the guided tour triggering methods for target exhibits can include exhibit QR code scanning, GPS positioning, near-field beacon, and a combination of GPS positioning and near-field beacon. Regardless of the triggering method used, the system queries the currently effective version based on the exhibit identifier and makes the guided tour resources subject to the main version number of the currently effective version.

[0059] The cache synchronization control module is used to compare the local cache version number with the main version number of the content currently in effect, and when the local cache version number is inconsistent with the main version number of the content, it sends a cache invalidation command and navigation resources to the WeChat Mini Program.

[0060] Specifically, the cache synchronization control module first queries the currently active version based on the exhibit identifier and language identifier, and reads the main version number of the content in the currently active version. Then, it compares the local cache version number uploaded by the WeChat Mini Program with the main content version number. If the local cache version number matches the main content version number, a cache availability flag is returned to the WeChat Mini Program, allowing the WeChat Mini Program to directly use the locally cached explanatory text, audio resources, and image resources. If the local cache version number does not match the main content version number, a cache invalidation command is returned to the WeChat Mini Program.

[0061] Furthermore, the cache synchronization control module is also used to obtain the hash value of local resources uploaded by the WeChat Mini Program; compare the local resource hash value with the resource hash value in the currently effective version to determine the difference resources; if the difference resources meet the differential update conditions, then send the differential resource package to the WeChat Mini Program; if the difference resources do not meet the differential update conditions, then send the complete resource package to the WeChat Mini Program.

[0062] Differential update conditions may include the number of differential resources being less than a preset number, the total size of differential resources being less than a preset size, mandatory upgrades to the main version number of content not involved, and network conditions meeting differential transmission requirements. A differential resource package may only include updated target language explanation text, updated audio resources, or updated image resources; a complete resource package may include all explanation text, audio resources, image resources, and AI question-and-answer retrieval entry information corresponding to the currently effective version.

[0063] The AI ​​Q&A output module is used to respond to AI Q&A requests sent by WeChat Mini Programs. It limits the scope of AI Q&A knowledge fragment retrieval based on exhibit identifiers, language identifiers, current effective versions, and permission tokens, and generates guide answers based on the retrieved AI Q&A knowledge fragments.

[0064] Specifically, after a user inputs or voices a question about a target exhibit in the WeChat mini-program, the mini-program sends an AI Q&A request to the AI ​​Q&A output module. The AI ​​Q&A request includes the question text, exhibit identifier, language identifier, current version, and permission token. The AI ​​Q&A output module first verifies the permission token to determine if the user has permission to access the target exhibit, the language, and the AI ​​Q&A function. After successful permission verification, the AI ​​Q&A output module only searches for AI Q&A knowledge fragments that match the exhibit identifier, language identifier, current version, and scope of permission.

[0065] If a matching AI question-and-answer knowledge fragment is found, the AI ​​question-and-answer output module inputs the question text and the retrieved AI question-and-answer knowledge fragment into the large language model to generate a guide-based answer corresponding to the language identifier. If no matching AI question-and-answer knowledge fragment is found, the AI ​​question-and-answer output module returns a confirmation message without using AI question-and-answer knowledge fragments that are not currently active to generate an answer.

[0066] By using the above methods, the AI ​​question-and-answer output module can avoid calling up knowledge fragments from historical versions, reducing the possibility of inconsistencies between the AI's answers and the current exhibit explanation text.

[0067] Example 2 This embodiment provides a multilingual intelligent navigation method based on WeChat Mini Programs, used in the multilingual intelligent navigation system based on WeChat Mini Programs in Embodiment 1, such as... Figure 2 As shown, the method includes the following steps: S1. Receive the exhibit content update request for the target exhibit and generate candidate versions of the exhibit content.

[0068] Specifically, the cloud server receives exhibit content update requests from the backend administrator for target exhibits. These requests can include adding target exhibits, modifying source language explanation text, modifying target language explanation text, replacing audio resources, replacing image resources, modifying extended knowledge text, or adjusting the scope of applicable permissions. The cloud server encapsulates this update content into candidate versions of exhibit content and generates candidate version identifiers for these candidate versions.

[0069] S2. Perform semantic unit matching and key entity matching on the source language explanation text and the target language explanation text to generate content consistency evaluation results.

[0070] Specifically, the system performs semantic unit decomposition on the source language explanation text and the target language explanation text respectively to obtain source language semantic units and target language semantic units; maps the source language semantic units and target language semantic units to a unified semantic space, calculates the semantic similarity between the two, and establishes a semantic matching relationship.

[0071] The system also extracts key entities from the source language and target language explanation texts, standardizes these key entities, and generates key entity fingerprints. The system compares the key entity fingerprints corresponding to the source language and target language explanation texts to identify instances of entity consistency, missing entities, and entity conflicts.

[0072] Simultaneously, the system checks the resource reference integrity among the target language explanatory text, audio resources, image resources, and extended knowledge text. Based on semantic matching relationships, entity matching relationships, and resource reference integrity, the system generates content consistency evaluation results.

[0073] S3. When the content consistency evaluation results meet the release conditions, generate AI question-and-answer knowledge fragments and bind exhibit identifiers, language identifiers and candidate version identifiers to the AI ​​question-and-answer knowledge fragments.

[0074] Specifically, if the content consistency evaluation results meet the publication conditions, the system will segment the source language explanation text, the target language explanation text, and the extended knowledge text into candidate knowledge fragments. Semantic deduplication and entity completion will then be performed on these candidate knowledge fragments to obtain AI question-and-answer knowledge fragments. Each AI question-and-answer knowledge fragment is bound to an exhibition hall identifier, exhibit identifier, language identifier, candidate version identifier, main content version number, and applicable permission scope.

[0075] If the content consistency evaluation result does not meet the release conditions, the system will not generate currently publishable AI question-and-answer knowledge fragments and will mark the candidate version of the exhibit content as pending correction.

[0076] S4. Generate an exhibit content version package based on the candidate versions of the exhibit content, and set the exhibit content version package as the currently effective version.

[0077] Specifically, the system generates a main version number for candidate versions of exhibit content and determines separate version numbers for text resources, audio resources, image resources, and AI question-and-answer knowledge fragments. The system then writes the main version number, text resource version number, audio resource version number, image resource version number, AI question-and-answer knowledge fragment version number, resource hash value, key entity fingerprint, and scope of permissions into the same exhibit content version package.

[0078] When the text resources, audio resources, image resources, and AI Q&A knowledge fragments in the exhibit content version package all meet the release conditions, the system will set the exhibit content version package as the currently effective version and mark the AI ​​Q&A knowledge fragments corresponding to the previous versions as not currently effective.

[0079] S5. Respond to the WeChat mini program's request to scan the exhibit's QR code, and obtain the exhibit identifier, language identifier, local cache version number, and permission token.

[0080] Specifically, after a user scans the exhibit's QR code using a WeChat mini-program, the mini-program parses or uploads the QR code encoding information. The cloud server obtains the exhibition hall identifier, exhibit identifier, and QR code batch identifier, and determines whether the exhibit's QR code belongs to a valid QR code batch based on the batch identifier. If the exhibit's QR code belongs to a valid batch, the system queries the currently active version based on the exhibit identifier. The WeChat mini-program simultaneously uploads the user-selected language identifier, the locally cached version number, and the permission token.

[0081] In other implementations, the WeChat mini-program can also obtain GPS location information and near-field beacon trigger information. The system determines the user's exhibition area based on the GPS location information, identifies the target exhibit based on the near-field beacon trigger information, and generates a guided tour trigger request for the target exhibit when the user's exhibition area matches the exhibition area information of the target exhibit. The guided tour trigger request includes exhibit identifier, language identifier, local cache version number, and permission token. After receiving the guided tour trigger request, the system continues to compare the local cache version number with the current effective version's main content version number according to step S6.

[0082] S6. Compare the local cache version number with the main content version number of the currently effective version, and issue a cache invalidation command and navigation resources when the local cache version number is inconsistent with the main content version number.

[0083] Specifically, the system reads the major version number of the currently active content and compares it with the local cache version number. If the local cache version number matches the major content version number, the system returns a cache availability flag to the WeChat Mini Program. If the local cache version number does not match the major content version number, the system issues a cache invalidation command to the WeChat Mini Program.

[0084] The system further compares the hash values ​​of local resources uploaded by the WeChat mini-program with the hash values ​​of resources in the currently effective version to determine the differing resources. If the differing resources meet the differential update conditions, a differential resource package is issued; if the differing resources do not meet the differential update conditions, a complete resource package is issued. The navigation resources include explanatory text, audio resources, image resources, and AI question-and-answer retrieval entry information in the corresponding language.

[0085] S7. Respond to the AI ​​Q&A request sent by the WeChat mini program, limit the scope of AI Q&A knowledge fragment retrieval based on exhibit identifier, language identifier, current effective version and permission token, and generate guide answer content.

[0086] Specifically, after a user enters a question in the WeChat mini-program or inputs a question via voice, the mini-program sends an AI Q&A request. The system uses an access token to determine whether the user has permission to access the current target exhibit, the current language, and the AI ​​Q&A service. Once the permission verification is successful, the system retrieves AI Q&A knowledge fragments based on exhibit identifier, language identifier, current version, and scope of permission application.

[0087] The system inputs the retrieved AI Q&A knowledge fragments and user questions into a large language model to generate guided tour responses. The language of the guided tour responses matches the language identifier. The system can also return the exhibit identifier, main version number, AI Q&A knowledge fragment version number, and knowledge fragment number used to generate the guided tour responses to the WeChat mini program for backend traceability.

[0088] Example 3 This embodiment uses a target exhibit from a museum as an example to illustrate the above system and method.

[0089] The backend administrator changed the mid-Qing Dynasty to the late Qing Dynasty in the Chinese source language description text of a target exhibit, and simultaneously modified the English target language description text for the same exhibit. Upon receiving this exhibit content update request, the content update management module generates candidate versions of the exhibit content.

[0090] The content consistency verification module breaks down the Chinese source language explanation text into multiple source language semantic units and the English target language explanation text into multiple target language semantic units, and calculates semantic matching relationships. Simultaneously, the system extracts key entities such as the name and type of the late Qing Dynasty target exhibits, and standardizes key entities such as the English names and types of the late Qing Dynasty target exhibits in the English target language explanation text, generating key entity fingerprints.

[0091] If the English target language explanation text remains as "middleQingdynasty", the key entity fingerprint is inconsistent with the source language explanation text, and the content consistency evaluation result does not meet the publication conditions. The system marks the candidate version of the exhibit's content as pending correction and generates a prompt message, reminding the administrator to verify the historical date entities.

[0092] After the administrator corrected the English target language explanation text to lateQingdynasty, the content consistency evaluation result met the release conditions. The system regenerated the English audio resource, calculated the resource hash value of the English audio resource, and bound it to the text resource version number of the English target language explanation text. The AI ​​question-answering index synchronization module generated AI question-answering knowledge fragments based on the corrected English target language explanation text and extended knowledge text, and bound each AI question-answering knowledge fragment to an exhibit identifier, language identifier, and candidate version identifier.

[0093] The content version package generation module writes Chinese text, English text, Chinese voice resources, English voice resources, image resources, AI question-and-answer knowledge fragments, resource hash values, key entity fingerprints, and scope of permissions into the same exhibit content version package, and sets the exhibit content version package as the currently effective version.

[0094] When a user scans the QR code of a target exhibit via a WeChat mini-program, the scanning access control module obtains the exhibit's identifier and receives the language identifier, local cache version number, and permission token uploaded by the WeChat mini-program. Alternatively, when a user enters the exhibition area of ​​the target exhibit and receives near-field beacon trigger information bound to that exhibit, the scanning access control module determines the exhibit's identifier based on GPS location information and near-field beacon trigger information, and continues to perform current effective version query and cache synchronization. If the cache synchronization control module finds that the local cache version number in the WeChat mini-program is inconsistent with the current effective version, it issues a cache invalidation command and issues a differential resource package based on the resource hash value comparison result. When the user plays the English narration again, the WeChat mini-program plays a new audio resource bound to the current English target language narration text.

[0095] When a user asks about the age of a target exhibit in English, the AI ​​Q&A output module limits the scope of AI Q&A knowledge fragment retrieval based on the exhibit identifier, English language identifier, current version, and permission token. It only retrieves English AI Q&A knowledge fragments corresponding to the current version and generates English guided tour answers. Because AI Q&A knowledge fragments for historical versions have been marked as not currently in effect, the system will not reference old knowledge fragments corresponding to middleQingdynasty.

[0096] Example 4 This embodiment provides an electronic device, including a processor, a memory, and a communication interface, which are connected via a communication bus. The memory stores a computer program, which, when executed by the processor, implements a multilingual intelligent navigation method based on a WeChat mini-program, as described in Embodiment 2.

[0097] The processor can be a central processing unit, graphics processing unit, digital signal processor, or computing unit in a cloud server. Memory can include read-only memory, random access memory, solid-state drive, or cloud storage. Communication interfaces can include wired network interfaces, wireless network interfaces, Bluetooth interfaces, 5G communication interfaces, or network communication interfaces for connecting to the WeChat mini-program server.

[0098] Electronic devices can be deployed on cloud servers, local servers within the exhibition hall, or cloud-edge collaborative servers. These devices receive QR code scanning requests, AI Q&A requests, and cache synchronization requests from the WeChat mini-program via a communication interface, and return guide resources, cache invalidation instructions, differential resource packages, complete resource packages, and guide answer content to the WeChat mini-program.

[0099] Through the above implementation methods, the present invention enables the source language explanation text, target language explanation text, audio resources, image resources, AI question-and-answer knowledge fragments, and WeChat mini-program local cache resources of the target exhibits to be updated collaboratively under the same currently effective version, thereby reducing the problem of version inconsistency between different resources during multilingual intelligent tours.

[0100] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multilingual intelligent navigation system based on WeChat mini-programs, characterized in that, include: The content update management module is used to receive the exhibit content update request of the target exhibit and generate exhibit content candidate versions. The exhibit content candidate versions include exhibit identifier, source language explanation text, target language explanation text, audio resources, image resources and extended knowledge text. The content consistency verification module is used to perform semantic unit matching and key entity matching on the source language explanation text and the target language explanation text to generate a content consistency evaluation result. The AI ​​Q&A index synchronization module is used to generate AI Q&A knowledge fragments based on the candidate versions of the exhibit content when the content consistency evaluation results meet the release conditions, and to bind exhibit identifiers, language identifiers and candidate version identifiers to the AI ​​Q&A knowledge fragments. The content version package generation module is used to generate an exhibit content version package based on the candidate versions of the exhibit content, and set the exhibit content version package as the currently effective version. The exhibit content version package includes exhibit identifier, language identifier, audio resources, image resources, AI question and answer knowledge fragments, resource hash values, and scope of applicable permissions. The QR code access control module is used to respond to the WeChat mini program's request to scan the exhibit's QR code, obtain the exhibit identifier corresponding to the exhibit's QR code, and receive the language identifier, local cache version number, and permission token uploaded by the WeChat mini program. The cache synchronization control module is used to compare the local cache version number with the content main version number of the currently effective version, and when the local cache version number is inconsistent with the content main version number, it sends a cache invalidation instruction and navigation resources to the WeChat Mini Program. The AI ​​Q&A output module is used to respond to AI Q&A requests sent by WeChat Mini Programs. It limits the scope of AI Q&A knowledge fragment retrieval based on the exhibit identifier, language identifier, current effective version, and permission token, and generates guide answer content based on the retrieved AI Q&A knowledge fragments.

2. The multilingual intelligent navigation system based on WeChat mini-programs according to claim 1, characterized in that, The QR code access control module is also used to parse the QR code encoding information in the exhibit's QR code, which includes the exhibition hall identifier, exhibit identifier, and QR code batch identifier. Determine whether the exhibit's QR code belongs to a valid QR code batch based on the batch identifier of the QR code; If it belongs to a valid QR code batch, then query the currently effective version based on the exhibit identifier; If it does not belong to a valid QR code batch, a QR code expiration message or exhibit redirection information will be generated.

3. The multilingual intelligent navigation system based on WeChat mini-programs according to claim 1, characterized in that, The content consistency verification module includes: The semantic unit matching unit is used to split the source language explanation text into source language semantic units, split the target language explanation text into target language semantic units, and establish a semantic matching relationship between the source language semantic units and the target language semantic units. The key entity matching unit is used to extract key entities from the source language explanation text and the target language explanation text, and to establish entity matching relationships between key entities in the source language and key entities in the target language. The resource reference detection unit is used to detect the integrity of resource references among the target language explanatory text, audio resources, image resources, and extended knowledge text; The consistency evaluation unit is used to generate the content consistency evaluation result based on the semantic matching relationship, entity matching relationship, and resource reference integrity.

4. A multilingual intelligent navigation system based on WeChat mini-programs according to claim 3, characterized in that, The key entities include exhibit names, personal names, historical dates, place names, dynasties, types of artifacts, and quantity information; The key entity matching unit is also used to standardize the key entities, generate key entity fingerprints, and determine whether the source language explanation text and the target language explanation text correspond to the same exhibit semantic version based on the key entity fingerprints.

5. A multilingual intelligent navigation system based on WeChat mini-programs according to claim 4, characterized in that, The content version package generation module is also used to generate the main content version number, text resource version number, voice resource version number, image resource version number, and AI question-and-answer knowledge fragment version number; Write the main version number of the content, the version number of the text resource, the version number of the voice resource, the version number of the image resource, the version number of the AI ​​question-and-answer knowledge fragment, the resource hash value, the key entity fingerprint, and the scope of permission application into the same exhibit content version package.

6. A multilingual intelligent navigation system based on WeChat mini-programs according to claim 1, characterized in that, The AI ​​question-answering index synchronization module is also used to segment the source language explanation text, the target language explanation text, and the extended knowledge text into candidate knowledge fragments; Semantic deduplication and entity completion are performed on the candidate knowledge fragments to obtain AI question-answering knowledge fragments; An AI question-and-answer retrieval index is generated based on the AI ​​question-and-answer knowledge fragments, and the AI ​​question-and-answer knowledge fragments corresponding to historical versions are marked as not currently in effect.

7. A multilingual intelligent navigation system based on WeChat mini-programs according to claim 1, characterized in that, The cache synchronization control module is also used to obtain the hash value of local resources uploaded by the WeChat mini program; The local resource hash value is compared with the resource hash value in the currently effective version to determine the differing resources; If the differential resources meet the differential update conditions, then a differential resource package is sent to the WeChat mini program; If the differential resources do not meet the differential update conditions, a complete resource package will be sent to the WeChat mini program.

8. A multilingual intelligent navigation system based on WeChat mini-programs according to claim 1, characterized in that, The content update management module is also used to generate corresponding language audio resources based on the target language explanation text when the target language explanation text meets the publishing conditions, or to receive manually recorded audio resources corresponding to the target language explanation text. Calculate the resource hash value of the audio resource and bind the resource hash value of the audio resource to the text resource version number of the target language explanation text; When the version number of the text resource for the target language explanation text changes, the audio resource bound to the previous text resource version number is marked as pending update.

9. A multilingual intelligent navigation method based on WeChat Mini Programs, characterized in that, The method for a multilingual intelligent navigation system based on WeChat mini-programs as described in any one of claims 1 to 8 includes: S1. Receive the exhibit content update request for the target exhibit and generate candidate versions of the exhibit content; S2. Perform semantic unit matching and key entity matching on the source language explanation text and the target language explanation text to generate content consistency evaluation results; S3. When the content consistency evaluation result meets the release conditions, generate an AI question-and-answer knowledge fragment and bind the exhibit identifier, language identifier and candidate version identifier to the AI ​​question-and-answer knowledge fragment; S4. Generate an exhibit content version package based on the candidate versions of the exhibit content, and set the exhibit content version package as the currently effective version; S5. Respond to the WeChat mini program's request to scan the exhibit's QR code, and obtain the exhibit identifier, language identifier, local cache version number, and permission token; S6. Compare the local cache version number with the content main version number of the currently effective version, and issue a cache invalidation command and navigation resources when the local cache version number is inconsistent with the content main version number; S7. Responding to the AI ​​Q&A request sent by the WeChat mini program, the scope of AI Q&A knowledge fragment retrieval is limited based on the exhibit identifier, language identifier, current effective version and permission token, and guide answer content is generated.

10. An electronic device, characterized in that, It includes a processor, a memory, and a communication interface, which are connected via a communication bus. The memory stores a computer program, which, when executed by the processor, implements the multilingual intelligent navigation method based on WeChat mini-programs as described in claim 9.

Images